Deductive Entropy and the Evolution of Uncertainty Filtering: A Unified Computational Framework for Autism Spectrum Disorder and Psychosis

Background: Autism Spectrum Disorder (ASD) and psychosis, despite their apparent divergence, share core disruptions in the brain’s ability to filter uncertainty. I introduce the concept of deductive entropy —the process by which the brain unconsciously resolves uncertainty, permitting only sufficiently certain predictions to enter consciousness—and demonstrate its utility as a computational and translational model for both conditions. Methods: A Monte Carlo simulation model was constructed, reflecting developmental trajectories, neurobiological parameters, timing of filtering breakdown, and adaptive learning. The model’s performance was benchmarked against epidemiological, clinical, and neurobiological data for ASD and psychosis. I evaluated the effect sizes and likelihood of benefit for key therapeutic interventions targeting uncertainty filtering, adaptive compensation, and early prevention. Results: The deductive entropy model provides a robust explanatory framework for the neurobiological processes underlying both ASD and psychosis. Simulation results demonstrate that early, persistent disruption of uncertainty filtering closely matches observed features of ASD (AUC = 0.94, 95% CI: 0.91–0.97; PPV 0.91, NPV 0.90). For psychosis, a late breakdown after adaptive learning history recreates core clinical features (AUC = 0.87, 95% CI: 0.82–0.93). Importantly, these metrics validate the model’s explanatory—not diagnostic—power, confirming that phenotypic differences are emergent properties of timing and adaptation within a unified neurobiological process. Therapeutically, polytherapy and uncertainty-targeted early intervention yielded the highest likelihood of clinical benefit in ASD, while executive/metacognitive and salience-modulating therapies were most effective in psychosis. Conclusions: Deductive entropy offers a unified, neurobiologically grounded model of ASD and psychosis, integrating predictive coding, adaptive learning, and filtering failure. The results provide a roadmap for precision, developmentally timed, and uncertainty-targeted interventions across the neurodevelopmental–psychosis spectrum. This work exemplifies translational psychiatry by integrating computational neuroscience, clinical psychiatry, and therapeutic modelling. By introducing deductive entropy as a recursive, threshold-based gate across inference layers, the model also offers a testable neurobiological account of consciousness and psychiatric dysfunction, with broad relevance to ASD, psychosis, and beyond.